The US has a few of non-combat coded F-22 in training brigades, those were mostly the oldest production, and I believe some of the LRIP aircraft as well.
Those airframes are old though, at least twice as old as the oldest non-prototype J-20 airframe.
And besides, its the US that still holds quantitative advantage in stealth fighters (overall, as logistics constrains them for a localised Pacific conflict, but this is a malleable number). F-22 airframes would therefore have lower priority than J-20 airframes, not to mention its not even the main fighter the US expects to rely on for a Pacific conflict anymore due to its lack of modern 5th-gen capabilities (e.g., sensor fusion). The J-20 is still very much expected to be the main fighter for a Pacific conflict.
J-20 WS-10C already dominates J-20 AL-31 in exercises like Golden Helmet. Imagine how big a gap J-20A/S has with vanilla J-20 produced a decade ago. With the sheer number of fifth gen aircraft PLAAF is cranking out, they can afford to put the best steel where they are needed.
That said AL-31 J-20 can be invaluable in training fifth gen combat philosophy. There is already a huge combat philosophy change between 4.5th gen and 4th gen fighter pilots so imagine how bad it would be to switch from 4th gen to 5th gen. Better teach them the right habits rather than let them unlearn bad habits when they start performing interception missions in the East China Sea.
I had the same thoughts, and its not that I disagree. The real question is the magnitude of the difference. Is it worth sacrificing a few of the oldest and weakest airframes of the far-and-away superior generation of fighters (that are still useful in peer conflicts and likely dominant in all other conflicts) to be full-time trainers? And how many airframes?
I don't know if we'll ever get enough details to make that decision, but that's fine since its not our decision to make. But this is a good opportunity to get insight into PLA's thoughts on J-20. If we do see some combat-capable J-20s be relegated to full training even while total J-20 family numbers are 300-400 while F-35 is >1000, that suggests they feel confident enough in reducing some overall fleet capability today to train for greater capability a little later.
There's also the question of if the gap between AL-31 J-20 and WS-10C2/WS-15 J-20A is so great that habits learned on the former would be a detriment to just having a clean learning approach to the latter.